JOURNAL OF MOLECULAR SPECTROSCOPY ARTICLE NO.

186, 422447 (1997)

MS977449

High-Temperature Rotational Transitions of Water in Sunspot and Laboratory Spectra

Oleg L. Polyansky,* ,1 Nikolai F. Zobov,* ,1 Serena Viti,* Jonathan Tennyson,* Peter F. Bernath, ,2 and Lloyd Wallace
*Department of Physics and Astronomy, University College London, Gower Street, London WC1E 6BT, United Kingdom; Departments of Chemistry and Physics, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1; and National Optical Observatories, P.O. Box 26732, Tucson, Arizona 85732 Received August 22, 1997

Assignments are presented for spectra of hot water obtained in absorption in sunspots (T 3000 C and 750 n 1010 cm01 ) and in emission in the laboratory (T 1550 C and 370 n 930 cm01 ). These assignments are made using variational nuclear motion calculations based on a high-level ab initio electronic surface, with allowance for both adiabatic and nonadiabatic corrections to the BornOppenheimer approximation. Some 3000 of the 4700 transitions observed in the laboratory spectrum are assigned as well as 1687 transitions observed in the sunspot spectrum. All strong lines are now assigned in the sunspot measurements. These transitions involve mostly high-lying rotational levels within the (0,0,0), (0,1,0), (0,2,0), (1,0,0), and (0,0,1) vibrational states. Transitions within the (0,3,0), (0,4,0), (1,1,0), (0,1,1), (0,2,1), (1,1,1), (1,2,0), and (1,0,1) states are also assigned. For most bands the range of Ka values observed is significantly extended, usually doubled. New features observed include numerous cases where the closely degenerate levels JKaKc and JKaKc/1 with high Ka are split by Coriolis interactions. Comparisons are made with the recent 1997 Academic Press line list of Partridge and Schwenke (1997, J. Chem. Phys. 106, 4618).
I. INTRODUCTION

The infrared spectrum of water is arguably the most important of all molecules because there are a multitude of applications. However, the spectrum of water is very complicated and the spectrum of hot water remains poorly understood. Detailed, line-by-line data on hot water are required for radiative transport models of many hot systems. Such systems include the spectra of oxygen-rich late-type stars (13) for which water vapor is the most important source of infrared opacity and substellar objects, such as brown dwarfs, for which water is the most abundant molecule after hydrogen (4, 5). Hot water is also one of the primary products of the combustion of hydrocarbons and has been detected in emission from forest fires (6) and from flames, for example, from an oxyacetylene torch (7). There are numerous military applications, including the simulation of rocket plumes (8) and the identification of ships, aircraft, helicopters, and tanks from their exhaust signatures. Recently spectra of hot water have been observed in sunspots (911) and at lower temperatures in the laboratory
Supplementary data for this article may be found on the journal home page (http://www.apnet.com/www/journal/ms.htm or http://www. europe.apnet.com/www/journal/ms.htm). 1 Permanent address: Institute of Applied Physics, Russian Academy of Science, Uljanov Street 46, Nizhnii Novgorod, Russia 603024. 2 Also: Department of Chemistry, University of Arizona, Tucson, AZ 85721.
0022-2852/97 $25.00 Copyright 1997 by Academic Press All rights of reproduction in any form reserved.

(12, 13). The sunspot spectra were originally considered unassignable and only small portions of the laboratory spectra have been assigned on a case-by-case basis (1214). However, in a recent paper (15), henceforth I, we demonstrated that by using a combination of high-level ab initio calculation and careful spectral analysis it was possible to assign a large number of new transitions in both these spectra. This assignment procedure represents a significant shift away from traditional, perturbation theory-based, methods of spectral assignment. In this paper we present full results of this analysis.
II. OBSERVED SPECTRA

Both the spectra under analysis have been published elsewhere and here we confine ourselves to brief details. The sunspot spectrum was published in an atlas by Wallace et al. (911) in the wavenumber range 420 n 1233 cm01 . It was recorded using the 1-m Fourier transform spectrometer of the McMathPierce Solar Telescope of the National Solar Observatory on Kitt Peak at a resolution of about 0.005 cm01 . At both ends of this spectrum there are significant gaps due to telluric absorption and some portions show many strong absorptions due to other molecules, such as SiO. However, the region 750 n 1010 cm01 is dominated by absorption due to water and we concentrated on this region. The observed sunspot spectrum is highly congested with

422

HOT ROTATIONAL TRANSITIONS OF H2O

423

up to 50 lines per wavenumber. Prior to I, there were no assigned water transitions in the spectrum but many transitions could be attributed to water on the basis of a comparison with the laboratory spectrum described below. The assigned SiO vibrationrotation lines in the sunspot spectrum suggest a temperature of T 3200 K (9, 10) and this temperature is assumed for the water lines. The laboratory spectrum was obtained to confirm that the sunspot features were indeed due to water (9). The emission spectrum, details of which can be found in Polyansky et al. (12), was recorded using a Bruker Fourier transform spectrometer located at the University of Waterloo (16). It covers the wavenumber range 370 n 930 cm01 and was recorded at a temperature of 1550 C. This spectrum has a similar resolution ( dn 0.01 cm01 ) to the sunspot spectrum but is much less crowded because of the lower temperature. While many of the water lines in the sunspot spectrum are blended, the lines in the laboratory spectrum are nearly all resolved.
III. VARIATIONAL CALCULATIONS

Assignments were made using synthetic spectra generated from an ab initio line list. Comparisons could also be made with recently published line lists (1719) based on calculations made with effective, spectroscopically determined potential energy surfaces. We note that, in general, these fitted potentials give line positions which have significantly smaller absolute errors than our ab initio line list, but that the errors in the ab initio line list are much more systematic. This smooth behavior of the errors proved to be crucial for making reliable assignments. The use of more than one line list is helpful for checking assignments. Similar remarks were made in the course of the recent analysis of the spectrum of H / (20); 3 however, the spectroscopically determined H / line list repro3 duced the experimental data to much higher accuracy ( 0.015 cm01 ) than any of the line lists available for water. A more detailed comparison of the line lists derived from fitted and ab initio potentials is given below. Energy levels and wavefunctions were generated for our line list using a very high-quality ab initio (BornOppenheimer) potential energy surface due to Partridge and Schwenke (21). Allowance for non-BornOppenheimer effects was made by adding an ab initio mass-dependent adiabatic surface (22) and by adjusting the effective atomic masses to allow, in part, for nonadiabatic effects. Following Zobov et al. (22) we used an H atom mass of 1.007551 amu, midway between that of H and a proton, and an O atom of 15.990526 amu. Dipole transitions were calculated using the ab initio dipole surfaces of Gabriel et al. (23). Calculations were performed using the DVR3D program suite (24) which was modified somewhat to improve performance for large calculations such as these. The calculations were performed using symmetrized Radau coordinates (r1 , r2 , u ) (25), which for water are close to, but more computaCopyright

tionally convenient than, bond lengthbond angle coordinates. We used a DVR grid of 40 points based on Gauss (-associated) Legendre polynomials in the u coordinate. For the radial coordinates we used a DVR grid of 21 points with radial basis set parameters of re 2.06a0 , De 0.14Eh , and ve 0.014Eh , where re is the equilibrium radius, De is the dissociation energy, and ve is the fundamental frequency of the Morse oscillator-like functions upon which the DVR is based ( 24). This number of grid points is sufficient to obtain good convergence for low-lying vibrational levels (26). In the first vibrational step we diagonalized a series of final secular problems of dimension 1000 from which we retained the lowest 500 eigenvalues and eigenvectors. For given J , the full rovibrational problem was solved using a basis of the 150 1 ( J / 1 ) lowest solutions from the first step. The number of eigenvalues obtained varied with J with no fixed rule. We computed eigenvalues with energies of up to at least 18 000 cm01 for J 25; for J 25 this corresponds to the lowest 500 eigenvalues for each symmetry block. For 25 J 33, eigenenergies up to 23 000 cm01 were computed corresponding to 320 eigenvalues per symmetry block for J 33. These criteria ensured that we covered all the energy levels belonging to the (000), (100), (010), (020), (001), and (030) vibrational states and of course, low Ka states of many higher vibrational bands are also included. Temperature-dependent spectra were generated using an adapted version of program SPECTRA ( 27) which uses full nuclear spin statistics and only rigorous dipole transition selection rules.
IV. SPECTRAL ANALYSIS

IV.1. Procedure For both laboratory and sunspot spectra, assignments were made in a series of steps. First, all transitions between energy levels known from previous work ( 7, 12, 13, 2832) were assigned. Such assignments, which we will call trivial, form only a minor part of both spectra. The next step was to use our line list to generate synthetic spectra at an appropriate temperature. For the laboratory spectra, intensities were observed to vary from 0.7 to 0.003 in relative units ( 12). Synthetic spectra gave excellent line-by-line agreement in intensity over the entire dynamic range. This meant that intensities as well as line positions could be used as guides for assignment. For the sunspot the situation is more complicated. Without implementing a detailed radiative transport model it is not possible to use the measured absorptions directly to give line intensities. Instead the sunspot spectrum was divided into four echelons on the basis of the strength of these absorptions. Strong transitions show 6 8% absorption, medium transitions show 4 6% absorption, weak ab-

1997 by Academic Press

424

POLYANSKY ET AL.

TABLE 1 Assigned Line Frequencies (in cm01 ) of the Hot Water Spectrum Observed in Sunspots (Laboratory Frequencies Are Also Given where Observed)

Copyright

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

425

TABLE 1 Continued

Copyright

1997 by Academic Press

426

POLYANSKY ET AL.

TABLE 1 Continued

Copyright

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

427

TABLE 1 Continued

Copyright

1997 by Academic Press

428

POLYANSKY ET AL.

TABLE 1 Continued

Copyright

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

429

TABLE 1 Continued

Copyright

1997 by Academic Press

430

POLYANSKY ET AL.

TABLE 1 Continued

Copyright

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

431

TABLE 1 Continued

Copyright

1997 by Academic Press

432

POLYANSKY ET AL.

TABLE 1 Continued

Copyright

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

433

TABLE 1 Continued

Copyright

1997 by Academic Press

434

POLYANSKY ET AL.

TABLE 1 Continued

Copyright

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

435

TABLE 1 Continued

Copyright

1997 by Academic Press

436

POLYANSKY ET AL.

TABLE 1 Continued

Copyright

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

437

TABLE 1 Continued

Sunspot frequencies are from (11) and have an estimated absolute accuracy of 0.002 cm01. Laboratory frequencies given to 5 decimal places are from (12) and have an estimated absolute accuracy of 0.001 cm01. Laboratory frequencies given to 4 decimal places are derived from experimental term values.

sorptions show 2 4% absorption, and there is a great deal of poorly resolved structure at the less than 2% absorption level. Comparison with our synthetic spectra suggest that going from the strong to medium echelons represents a drop in transition intensity by a factor of about 5. In this
Copyright

work we present assignments to all of the strong transitions, nearly all of the medium transitions, and some of the weak lines. We are confident that the extension of our calculations to higher vibrational states will lead to the assignment of the remaining weaker transitions.

1997 by Academic Press

438

POLYANSKY ET AL.

A crucial step in our assignment procedure was the use of branches. For the purposes of this analysis, a branch was defined as a series of transitions with Ka J 0 na , where na is constant for each branch. In the laboratory spectrum, which contains a significant number of transitions involving states of low Ka , we also followed branches defined by Kc J 0 nc for a constant nc . Our variational calculations are not completely accurate, largely due to residual errors in the BornOppenheimer potential, but they give errors which vary slowly and systematically with J for a given branch. Typically, using our ab initio line list, the next member of a branch could be predicted with an accuracy of 0.02 cm01 . It is thus possible to step along members of the branch starting from low J transitions, which are generally well characterized, to the previously unassigned higher J lines. For transitions observed both in the laboratory and in sunspots, a crude confirmation of each assignment was obtained by comparing the ratio of line intensities with that estimated from Boltzmann distributions at the appropriate temperature. After assigning all strong and many medium lines in the sunspot spectrum by analyzing branches with high Ka , we were left with a set of approximately 50 unassigned mediumstrength lines. These transitions were assigned to states with intermediate values of Ka and the assignments were checked by comparison with the line lists generated using spectroscopically determined potentials (1719). We found the line list of Partridge and Schwenke (19), which became available after the bulk of the assignments reported here had been made, particularly useful for this purpose. IV.2. Results Table 1 presents the lines assigned in the sunspot spectrum. There are 1687 transitions, all of which, with the exception of the rotational difference transitions (13) discussed below, are pure rotational transitions which arise mainly from excited vibrational states. Transitions are assigned to the (000), (010), (020), (100), (001), (030), (110), (011), and (021) vibrational states. Table 2 which can be obtained from the electronic archive, contains the 4700 lines observed in the laboratory spectrum of Polyansky et al. (12), for which we have been able to make some 3000 assignments. Nearly all of these are rotational transitions involving the (000), (010), (020), (100), (001), (030), (040), (110), (011), (021), (111), (120), and (101) vibrational states; there are also a few P-branch transitions from the (0,1,0) bending fundamental. To illustrate both the success of our assignment procedure and some of the interesting features of the spectra, we have plotted two portions of both spectra labeled with assignments. Figure 1 shows the region 771.5775.5 cm01 in which transitions belonging to three distinct difference band transitions can be seen. Figure 2 shows the region 841.5849.5 cm01 in
Copyright

which two high Ka transitions which should appear as a single line are split by (presumably) Coriolis interactions with levels of a different vibrational state. It is not possible to plot the entire spectrum here. In paper I, we presented the region 870.0875.4 cm01 , in which many water transitions were observed in the laboratory, and the region 924.0925.6 cm01 , in which only one laboratory transition is observed. With a definitive and complete set of assignments it is possible to build up stacks of observed energy levels for water. Such tabulations are particularly useful for making trivial assignments in other spectral regions (see Ref. 33, for example). Table 3 presents the observed energy levels for the ground vibrational state. A full tabulation (Table 4) for the (000), (010), (020), (100), and (001) vibrational states has been placed in the electronic archive. These tabulations greatly extend the previous ones (7, 12, 29, 32), the results of which have been incorporated for completeness. These tabulations are provided only for the convenience of the reader and it should be stressed that the fundamental data are the assigned transitions not these derived energy levels. In particular nearly all of our newly derived energy levels are based on a single transition, which means that one misassignment leads to a large number of erroneous energy levels. In addition, the propagation of experimental errors degrades the accuracy of the derived energy levels. Our data are not complete enough to build similar energy level ladders for the higher vibrational states. Table 3 shows that we have approximately doubled the energy range of the rotational levels and obtained a large increase in the range of observed Ka levels.
V. DISCUSSION

The newly assigned transitions show two phenomena which are not observed in low-temperature spectra. The first of these is a family of so-called rotational difference bands that we have discussed previously (14, 15). These transitions are caused by the mixing of rotational manifolds of neighboring vibrational states via Fermi interactions. The mixing results in intensity sharing between pure rotational transitions and what are nominally vibrationrotation transitions. The perturbations cause a doubling of the number of rotational transitions from 2 to 4 (see Fig. 3 in Ref. 15). Differences between any three of these transitions give an exact prediction for the fourth transition, so that these quartets can be securely identified. Isolated interactions between two individual rovibrational levels are to be expected. However, the interesting feature of the difference transitions is that they are not caused by such a single interaction but by a whole series of perturbations, which give rise to a series, or bands of transitions. Thus the (100) (020) rotational difference band is strongest for rotational levels involving Ka 9 but is also observable for Ka 8, 10, and 11. In the course of the assignments reported here we also found rotational difference bands caused by

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

439

FIG. 1. Sunspot absorption spectrum (upper, Wallace et al. (9, 11) and laboratory emission spectrum (lower, Wallace et al. (9), Polyansky et al. (12)) in the wavenumber region 771.5775.5 cm01 . Newly assigned water transitions are labeled using standard notation JKaKc for rotational levels and ( n1n2n3 ) for vibrational levels.

(110) (030) and (120) (040) interactions with the same values of Ka as found in the (100) (020) case. The second unusual feature we observe is the localized doubling of a number of pure rotational transitions with high Ka . It is a standard result of rotational spectroscopy that levels with JKaKc and JKa/1Kc for high Kc and JKaKc and JKaKc/1 for high Ka are degenerate. For water this leads to transitions for J 25 which can be separated only at resolutions much higher than 0.001 cm01 . In the course of this work we found many isolated examples where these transitions are split, presumably by Coriolis interactions with rotational levels belonging to other nearby vibrational states. In the course of our variational calculations we have undertaken a number of comparisons with the line list and energy levels obtained by Partridge and Schwenke (PS) (19). PS optimized their ab initio potential using experimental line positions for low-lying rotational levels (J 5) taken from the HITRAN data base (34). In the calculation
Copyright

of their full line list PS truncated the basis set used for the variational calculations for J 4 at the level of their J 4 calculations. Note that one would expect more uniform convergence if they had used a final basis whose size was proportional to J / p, where the rotational or Wang parity is given by ( 01) J/p . For low J energy levels PSs calculations give superb results, reproducing experiment with a much higher accuracy than either the ab initio line list used here or Viti et al.s (VTP1) previous line list, which was generated using an empirical potential derived from spectroscopic data (35). However, for higher rotational states, particularly those with J 20, we find that a very high proportion of rotational states which one expects to be degenerate in fact show significant splittings in the PS line list. This phenomenon is not the same as the isolated splittings discussed above but seems to be a uniform property of the higher levels. Neither our analysis of the experimental levels, the present calculations, or the VTP1 line list shows any evidence for such splittings.

1997 by Academic Press

440

Copyright

POLYANSKY ET AL.

1997 by Academic Press

FIG. 2. Sunspot absorption spectrum (upper, Wallace et al. (9, 11)) and laboratory emission spectrum (lower, Wallace et al. (9), Polyansky et al. (12)) in the wavenumber region 841.5849.5 cm01 . Newly assigned water transitions are labeled using standard notation JKaKc for rotational levels and ( n1n2n3 ) for vibrational levels. Note that some of the stronger sunspot absorption features are due to other molecules, particularly OH (11).

HOT ROTATIONAL TRANSITIONS OF H2O

441

Copyright

1997 by Academic Press

FIG. 2 Continued

442

POLYANSKY ET AL.

Copyright

1997 by Academic Press

FIG. 2 Continued

HOT ROTATIONAL TRANSITIONS OF H2O

443

To illustrate this point, Table 5 gives a comparison for the observed J 24 rotational levels of the manifold of the (001) vibrational state. It is clear from Table 5 that PSs line list systematically gives splittings which appear to be artificial. In particular, for the high Ka levels, the p 1 parity rotational levels (those with Kc odd for this band) all lie below the p 0 parity (Kc even) levels with which they should be quasi-degenerate. Since PS truncated their variational rotationvibration calculations at 7500 energy-selected basis functions independent of the parity, p, this means that the p 1 calculations will contain states of higher cutoff energy that the p 0 calculation. Because of the variational principle, the p 1 states will be better converged and hence lower in energy. To test the hypothesis that the splitting found in PSs rotational levels is an artifact of a lack of variational convergence, we performed calculations using the spectroscopically optimized potential that PS used for their line lists. Our calculations were performed with the same basis set parameters that were used to construct our ab initio line list, as specified in Section III, and masses as specified by PS. For J 0, our calculations agree to within 0.01 cm01 with PSs up to 25 000 cm01 . However, for J 24 our calculations gave energy levels which were systematically lower than those of PS ( by up to 1 cm01 ) . Furthermore none of the artificial splittings are present in these new levels ( see Table 5 ) . This artificial splitting of lines has two possible consequences. The first is that it is more difficult to use the line list for assignments. The second consequence is perhaps more subtle. A major objective of PS was to construct a list of over 300 million water transitions in order to model the atmospheres of cool (T 20004000 K) stars. An important consideration in these radiative transport models is how the line absorptions fill in the gaps in the spectrum. Two transitions which, to within their linewidth, are coincident will have a rather different effect than two well-separated transitions. In the latter case the overall stellar opacity will be overestimated. For T 20004000 K, Boltzmann considerations suggest that transitions involving states with J 2030 are the dominant sources of opacity. Artificially doubling the number of lines for these J values could have serious consequences for the opacity prediction. Although we have not discussed our assignment procedure in great detail, there are, in fact, two aspects to making correct line assignments. One step is associating a particular transition with a particular combination of upper and lower energy levels; the other step is assigning quantum number labels to these two energy levels. These labels take two forms: rigorously exact quantum numbers or symmetries and approximate labels. For water the only exact quantum numbers are the total rotational angular momentum, J , the rotational parity, p , and the vibrational parity or ortho / para designation which we normally denote q ( 24 ) .
Copyright

The conventional vibrational quantum numbers ( n1 , n2 , n3 ) and the rotational labels ( Ka , Kc ) are only approximate quantum numbers, although their values are constrained by J , p , and q . Variational calculations are usually performed using the full symmetry of the system in question and therefore the calculated energy levels are automatically labeled with the exact quantum numbers. In this work, the other quantum numbers were determined as part of the procedure of following the branches in the spectrum. Although we have not implemented such an algorithm, it is attractive to have a method of automatically assigning approximate quantum numbers to the calculated energy levels. Indeed both PS and the recent HITEMP database (18) have implemented such procedures. A comparison with our manual assignments is interesting. In both cases we find that the reliability of these automatic assignments shows a strong dependence on energy, particularly of the vibrational level involved. The approximate quantum numbers given by HITEMP become untrustworthy for states above about 10 000 cm01 . PSs assignments are more reliable but they still give many wrong labels above about 15 000 cm01 . In both cases, where a level is mislabeled, the labeling given usually appears to be very wild; that is, it belongs to a level which is distant in energy to the one being considered. Furthermore neither dataset gives a well-constructed set of energy labels in the sense that there are occasional energy levels with the same labels and certain sets of labels which are completely absent. This results in a number of distinct transitions with identical labels. It should be stressed that we found no evidence for chaotic behavior in the states of water we analyze here. As far as we can tell, our assignments have yet to reach parts of the water spectrum where the energy level labels become ambiguous, except possibly in the case of some of the accidental degeneracies discussed above. / This is not the case for H 3 where the recorded high-resolution spectra already probe the region where such labeling becomes rather arbitrary ( 20 ) . Finally it should be noted that the sunspot spectrum is available at other wavelengths ( 10, 11, 36 ) . We have recently undertaken a detailed analysis of transitions lying in the so-called K band ( 33 ) with a particular focus on the region 4860 4930 cm01 where the water transitions are densest. This analysis commenced by using the energy levels determined here to make trivial assignments but resulted in assigning transitions between many states involving higher vibrational levels, not observed previously. These vibrational states are higher in energy than those probed here and a comparison with the energy levels of PS showed significant deviations between predictions and observations. Full results of this study will be presented elsewhere ( 33 ) .

1997 by Academic Press

444

POLYANSKY ET AL.

TABLE 3 Energy Levels of Water (in cm01 ) Obtained from the Wavenumbers of the Hot Water Spectra for the Vibrational Ground States

Copyright

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

445

TABLE 3 Continued

Copyright

1997 by Academic Press

446
TABLE 3 Continued

POLYANSKY ET AL.
/ bound H 3 molecule (20), which up until now has been something of a special case. This change will represent a major shift away from the traditional perturbation theorybased analysis of high-resolution spectra. This new method of analysis is computationally much more expensive but is not without its benefits. Here, when confronted with a vast quantity of unanalyzed and seemingly unassignable data, we have based our calculations on ab initio potentials, occasionally reinforced by the use of calculations based on spectroscopically determined poten-

TABLE 5 Energy Levels for J 24 Rotational Levels of the (001) State in cm01

VI. CONCLUSIONS

In this work we present detailed assignments for hot water spectra obtained in a sunspot and in the laboratory. Although the work presented here concerns hot water, many of the characteristics of the spectrum will be common to the spectra of other light polyatomic molecules at high temperatures. It would seem, therefore, that a full quantum mechanical treatment based on variational calculations and high-quality potential energy surfaces, possibly with allowance for the failure of the BornOppenheimer approximation, will be necessary for spectral analysis and modeling of such systems. Indeed this is already the norm for much simpler diatomic systems, which display none of the spectral features described above, and for van der Waals complexes (37). This procedure has also been applied to the chemically
Copyright

Note. Experimental levels with Ka 3 are from Flaud et al. (7); levels with Ka 11 are from this work. Calculated energy levels are from Partridge and Schwenke (19) (PS), from this work using PSs spectroscopically refined potential, and from this work, ab initio.

1997 by Academic Press

HOT ROTATIONAL TRANSITIONS OF H2O

447

tials. However, spectroscopically determined potentials, derived from iterative variational calculations and fits to known spectroscopic data, are already widely used for molecules such as water ( 18, 19, 35, 38 ) . The effective potentials derived from such calculations give a much more compact representation of the data than the traditional, vibrational-state by vibrational-state, perturbation theory-based approaches. They also have vastly superior extrapolation properties ( 35 ) . By the use of a full quantum mechanical treatment of the water nuclear motion problem we have managed to assign the seemingly unassignable high-temperature spectrum of water. Perturbation theory-based methods have proved to be excellent for interpreting the spectra of cool molecules. We suggest that a new spectroscopic paradigm is required for hot molecules based on variational calculations. This leaves superexcited molecules in the dissociation region as the final frontier for the theory of small-molecule rotationvibration spectroscopy.
ACKNOWLEDGMENTS
We thank H. Partridge and D. W. Schwenke for supplying their potential energy surface prior to publication. The authors acknowledge NATO Grant 5-2-05 / CRG951293 for making the experimental theoretical collaboration possible. N.F.Z. thanks the Royal Society for funding his visit to the University College London. The work of O.L.P. was supported in part by the Russian Fund for Fundamental Studies. This work was supported by the Natural Sciences and Engineering Research Council of Canada ( NSERC ) . Acknowledgment is made to the Petroleum Research Fund for partial support of this work. Support was also provided by the NASA Laboratory Astrophysics Program, the UK Engineering and Science Research Council, and the UK Particle Physics and Astronomy Research Council.

REFERENCES
1. K. H. Hinkle and T. G. Barnes, Astrophys. J. 227, 923934 (1979). 2. F. Allard, P. H. Hauschildt, S. Miller, and J. Tennyson, Astrophys. J. 426, L39L41 (1994). 3. H. R. A. Jones, A. J. Longmore, F. Allard, P. H. Hauschildt, S. Miller, and J. Tennyson, Mon. Not. R. Astron. Soc. 277, 767776 (1995). 4. T. Tsuji and K. Ohnaka, in Elementary Processes in Dense Plasmas (S. Ichimaru and S. Ogata, Eds.), p. 193, AddisonWesley, Reading, MA, 1995. 5. B. R. Oppenheimer, S. R. Kulkarni, K. Mathews, and T. Nakajima, Science 270, 14781479 (1995). 6. H. Worden, R. Beer, and C. P. Rinsland, J. Geophys. Res. 102, 1287 1299 (1997). 7. J.-M. Flaud, C. Camy-Peyret, and J.-P. Maillard, Mol. Phys. 32, 499 521 (1976). 8. NATO AGARD (Advisory Group for Aerospace Research and Development), Termenology and Assessment Methods of Solid Propellant Rocket Exhaust Signatures, Advisory Report 287, Propulsion and Energetics Panel, Working Group 21, Feb. 1993.

9. L. Wallace, P. F. Bernath, W. Livingston, K. Hinkle, J. Busler, B. Guo, and K. Zhang, Science 268, 11551158 (1995). 10. L. Wallace, W. Livingston, K. Hinkle, and P. F. Bernath, Astrophys. J. Suppl. Ser. 106, 165169 (1996). 11. L. Wallace, W. Livingston, and P. F. Bernath, An Atlas of the Sunspot Spectrum from 470 to 1233 cm01 (8.1 to 21 mm) and the Photospheric Spectrum from 460 to 630 cm01 (16 to 22 mm), NSO Technical Report 1994-01, Tucson, AZ, 1994. 12. O. L. Polyansky, J. R. Busler, B. Guo, K. Zhang, and P. Bernath, J. Mol. Spectrosc. 176, 305315 (1996). 13. O. L. Polyansky, N. F. Zobov, J. Tennyson, J. A. Lotoski, and P. F. Bernath, J. Mol. Spectrosc. 184, 3550 (1997). 14. O. L. Polyansky, J. Tennyson, and P. Bernath, J. Mol. Spectrosc., in press. 15. O. L. Polyansky, N. F. Zobov, J. Tennyson, S. Viti, P. F. Bernath, and L. Wallace, Science 277, 346348 (1997). 16. P. F. Bernath, Chem. Soc. Rev. 25, 111115 (1996). 17. S. Viti, J. Tennyson, and O. L. Polyansky, Mon. Not. R. Astron. Soc. 287, 7986 (1997). 18. L. S. Rothman, R. Gamache, J. W. Schroeder, A. McCann, and R. B. Wattson, SPIE Proc. 2471, 105 (1995); L. S. Rothman, private communication (1996). 19. H. Partridge and D. W. Schwenke, J. Chem. Phys. 106, 46184639 (1997). 20. B. M. Dinelli, L. Neale, O. L. Polyansky, and J. Tennyson, J. Mol. Spectrosc. 181, 142150 (1997). 21. D. W. Schwenke, private communication (1996). [Preliminary fit to the ab initio data of Ref. (19)] 22. N. F. Zobov, O. L. Polyansky, C. R. Le Sueur, and J. Tennyson, Chem. Phys. Lett. 260, 381385 (1996). 23. W. Gabriel, E.-A. Reinsch, P. Rosmus, S. Carter, and N. C. Handy, J. Chem. Phys. 99, 897900 (1993). 24. J. Tennyson, J. R. Henderson, and N. G. Fulton, Comput. Phys. Commun. 86, 175198 (1995). 25. J. Tennyson and B. T. Sutcliffe, Int. J. Quantum Chem. 42, 941952 (1992). 26. N. G. Fulton, Ph.D. thesis, Chap. 4, University of London (1994). 27. J. Tennyson, S. Miller, and C. R. Le Sueur, Comput. Phys. Commun. 75, 339364 (1993). 28. J.-M. Flaud and C. Camy-Peyret, J. Mol. Spectrosc. 51, 142150 (1974). 29. C. Camy-Peyret, J.-M. Flaud, J. P. Maillard, and G. Guelachvili, Mol. Phys. 33, 16411650 (1977). 30. J.-M. Flaud, C. Camy-Peyret, J.-P. Maillard, and G. Guelachvili, J. Mol. Spectrosc. 65, 219228 (1977). 31. C. Camy-Peyret, J.-M. Flaud, and J.-P. Maillard, J. Phys. Lett. 41, L23L26 (1980). 32. R. A. Toth, J. Opt. Soc. Am. B 10, 15261544 (1993). 33. O. L. Polyansky, N. F. Zobov, J. Tennyson, S. Viti, P. F. Bernath, and L. Wallace, Astrophys. J. 489, L205L208 (1997). 34. L. S. Rothman, R. R. Gamache, R. H. Tipping, C. P. Rinsland, M. A. H. Smith, D. C. Benner, V. Malathy Devi, J.-M. Flaud, C. Camy-Peyret, A. Perrin, A. Goldman, S. T. Massie, L. R. Brown, and R. A. Toth, J. Quant. Spectrosc. Radiat. Transfer 48, 55 (1992). 35. O. L. Polyansky, P. Jensen, and J. Tennyson, J. Chem. Phys. 101, 76517657 (1994). 36. L. Wallace and W. Livingston, An Atlas of a Dark Sunspot Umbral Spectrum from 1970 to 8640 cm01 (1.16 to 5.1 mm). NSO Technical Report, 1992-01, Tuscon, AZ, 1992. 37. J. M. Hutson, Annu. Rev. Phys. Chem. 41, 123154 (1990). 38. O. L. Polyansky, P. Jensen, and J. Tennyson, J. Chem. Phys. 105, 64906497 (1996).

Copyright

1997 by Academic Press